Build fails with SEXP_USE_MALLOC

[dpapavas]

This is really an issue, not a PR. It's not something I particularly need, but during the investigation of a bug I suspected to be due to the GC, I tried defining SEXP_USE_MALLOC, to see whether the behavior would change. This resulted in a couple of C compile errors, which I fixed, but building just fails a bit later when building Scheme library stubs. I didn't pursue the failure further, but I though I'd report it.

(Unless I'm missing something, this seems to be broken for quite a while - at least since 0.8, so if it has gone undetected until now, perhaps doing away with it altogether would be appropriate.)

[ashinn]

This option is a hack which isn't suitable to any non-trivial program, so I don't guarantee support for it.

If you're having trouble with GC I recommend SEXP_USE_BOEHM or SEXP_USE_CONSERVATIVE_GC at a first pass. There are other GC options to investigate further.

[dpapavas]

Thanks for the tips. I had tried to SEXP_USE_CONSERVATIVE_GC and it did indeed make the problem "go away", but I had already found out by then, that it was due to not preserving some values I should have been preserving (I suppose, because what needs to be preserved and what doesn't, isn't entirely clear).

I noticed you merged the PR. Does the source build for you correctly with it? As I mentioned, it only fixes the C side of the problem for me; I still get Scheme errors when the FFI processes certain library stubs.